Search

With the development of connected and automated vehicles, eco-driving control is reckoned to generate unprecedented potential on energy-saving in electrified powertrain. In this paper, a data-driven based eco-driving control strategy with efficient computation capacity is proposed for plug-in hybrid electric vehicles to achieve approximate optimal energy economy. An efficient hierarchical optimal control scheme is designed to mitigate the massive computational cost during velocity optimization and powertrain control. A data-driven optimal energy consumption cost model and an optimal battery current model are respectively constructed via two neural networks and served as the critic model and the system model during velocity optimization. Furthermore, the neural network-based dynamic programming is exploited to optimize the vehicle velocity by merging the data-driven models and Bellman optimality principle. The simulation results demonstrate that the proposed method can remarkably improve fuel economy by up to 16.7% in complicated driving conditions, compared with conventional sequential optimization methods. Furthermore, the data-driven control scheme can drastically improve the computational efficiency with slight sacrifice on fuel economy, compared with the optimum benchmark. • A data-driven based eco-driving control method is proposed. • Optimal energy consumption and battery current are modeled by two neural networks. • A neural network-based dynamic programming scheme is exploited. • The devised method obviously promotes fuel economy with high computation efficiency. [ABSTRACT FROM AUTHOR]

Eco-Driving has great potential in reducing the fuel consumption of road vehicles, especially under the connected and automated vehicles (CAVs) environment. Traditional model-based Eco-Driving methods usually require sophisticated models and cannot deal with complex driving scenarios. This paper proposes a hybrid reinforcement learning (RL) based Eco-Driving algorithm considering both the longitudinal acceleration/deceleration and the lateral lane-changing operations. A deep deterministic policy gradient (DDPG) algorithm is designed to learn the continuous longitudinal acceleration/deceleration to reduce the fuel consumption as well as to maintain acceptable travel time. Collecting the critic's value of each single lane from DDPG and integrating the information of adjacent lanes, a deep Q-learning algorithm is developed to make the discrete lane-changing decision. Together, a hybrid deep Q-learning and policy gradient (HDQPG) method is developed for vehicles driving along multi-lane urban signalized corridors. The method can enable the controlled vehicle to learn well-established longitudinal fuel-saving strategies, and to perform appropriate lane-changing operations at proper times to avoid congested lanes. Numerical experiments show that HDQPG can reduce fuel consumption by up to 46% with marginal or no increase of travel times. • Continuous state space for longitudinal accelerations and discrete state space for lateral lane changing. • A hybrid deep Q-learning and policy gradient (HDQPG) method is developed for Eco-Driving. • HDQPG can enable vehicles to learn well-established Eco-Driving strategies. • HDQPG can save fuel consumption by 46% compared with the baseline model. • HDQPG outperforms existing RL-based Eco-Driving methods. [ABSTRACT FROM AUTHOR]

With the rapid growth of urbanisation and associated vehicle numbers in China, urban transportation has become responsible for a large increase in the country's energy consumption and, consequently, its air pollution levels. Beijing has been making significant efforts to reduce transportation sector energy consumption and pollution. However, owing to the existence of a large number of private cars, the implementation of effective energy consumption and emission reduction measures has been difficult to achieve. This paper reviews the policies and regulations that have so far been developed to tackle Beijing's transportation sector issues. Perspective and analysis are given on the specific policies and regulations that have been employed, focusing on the overarching regulation mechanism, public transportation infrastructure, vehicle technology-related measures, vehicle activity-related measures and eco-driving -related measures, relevant experiences and lessons are summarised. A key finding concludes that the implementation of driving restriction policy can restrain pollutants emissions significantly, it shows energy consumption and emission reduction should not only be applied to public transportation, but the effective control of emissions from private cars should also attract more attention. To better control energy consumption of urban transport, various measures and solutions are proposed and discussed, including the further popularizing of eco-driving and the widespread application of e-mobility. Specific proposals for developing urban transportation policy are also discussed. [ABSTRACT FROM AUTHOR]

Existing optimal control systems for vehicles that consider the effect of road slopes use a cost function with fixed weights related to speed deviation, regardless of driving states on slopes. As a result, gravitational potential energy is not efficiently exploited and braking at down-slopes (which wastes energy) becomes unavoidable. Thus, there is still significant scope to improve fuel saving behavior on slopes. To address this opportunity, in this paper, we present a dynamic eco-driving system (EDS) for a (host) vehicle based on model predictive control (MPC) with fuzzy-tuned weights, which helps efficiently utilize the gravitational potential energy. In the proposed EDS, we formulate a nonlinear optimization problem with an appropriate prediction horizon and an objective function based on the factors affecting vehicle fuel consumption. The objective function's weight is tuned via fuzzy inference techniques using information of the vehicle's instantaneous velocity and the road slope angle. By considering the vehicle longitudinal dynamics, preceding vehicle's state, and road slope information (obtained from the digital road map), the optimization generates velocity trajectories for the host vehicle that minimizes fuel consumption and CO 2 emission. We also investigate the traffic flow performance of following vehicles (behind the host vehicle) in dense traffic; this was not considered in existing works on hilly roads. The effectiveness of the proposed EDS is evaluated using microscopic traffic simulations on a real road stretch in Fukuoka City, Japan, and the results demonstrate that the fuzzy-tuned MPC EDS significantly reduces fuel consumption and CO 2 emission of the host vehicle compared to the traditional driving (human-based) system (TDS) for the same travel time. In dense traffic, the fuel consumption and CO 2 emission of following vehicles are noticeably reduced. • A fast fuzzy-tuned MPC framework is developed for eco-driving on hilly roads. • A dynamically tunable cost function is introduced to optimize the use of gravity. • The framework is tested in dense traffic using data of a road in Fukuoka, Japan. • The controlled and following cars show improved fuel consumption and CO 2 emission. [ABSTRACT FROM AUTHOR]

• Majority of traditional fuel consumption models are impractical for eco-driving. • ANN outperform SVR in time-series modeling of fuel consumption. • Internal impact of speed, acceleration, and grade interactions is captured by ANN. • Cascaded use of ANNs with different complexity significantly improves the accuracy. • Class- and technology-specific models lead to higher average accuracy in a fleet. Many of the existing studies on vehicular fuel consumption estimation are criticized in aspects such as ignoring real-world training data, low diversity of test fleet, impracticality of models in real-world applications (i.e. instrument-independent eco-driving), or their prediction power in the non-linear multi-dimensional space of fuel consumption estimation. In this paper, we proposed a machine learning modeling method using large on-road data collected from a fleet of 27 vehicles. The usability of models in absence of specialized instruments was in focus. We tried to improve the accuracy of our base models by introducing engine-speed estimates through a cascaded modeling procedure. As a result, the accuracy of models reached 83%, while improvements as high as 37% were achieved depending on the technique (support vector regression or artificial neural networks) and vehicle class. Finally, we took the first step from vehicle-specific models towards category-specific modeling by a categorical analysis over fleet attributes. [ABSTRACT FROM AUTHOR]

• A bi-level MPC-based eco-driving strategy for CAHEVs is proposed. • The control strategy uses real-time traffic information via V2V and V2I. • The driving scenario classifier is designed for driving in mixed driving scenarios. • The strategy is tested in a realistic traffic simulation environment. • The results are compared to the conventional rule-based strategy. This paper proposes a bi-level eco-driving control strategy for connected and automated hybrid electric vehicles (CAHEVs) under mixed driving scenarios. First, the hybrid electric vehicle powertrain is modelled, and the communications via Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) are introduced as the main data sources for the decision-making of the control system. Next, the problem is divided into three objectives, namely, (1) safe driving, (2) energy management, and (3) exhaust emission reduction. Based on the real-time road information, the driving scenario classifier (DSC) works towards determining the corresponding vehicle mode on which the cost function can be adjusted accordingly. The simulation is carried out in a realistic urban traffic simulation environment in SUMO. The results show that with the proposed model predictive control (MPC)-based strategy applied, safe driving in a trip involving a mixture of driving scenarios can be guaranteed throughout the entire driving. In addition, in comparison to the rule-based benchmark strategy, the proposed strategy can reduce the fuel consumption by 34.10% with battery kept in a healthy state of charge range, and the exhaust emissions (HC, CO, and NOx) are reduced by 25.36%, 72.30%, and 30.39%, respectively, which demonstrates the effectiveness and robustness of the proposed MPC-based strategy for CAHEVs. [ABSTRACT FROM AUTHOR]

• A fuel consumption model for Hybrid Electric Vehicles (HEVs) is developed. • Real driving tests of Argonne National Laboratory are used to verified the model. • HEVs show considerable fuel saving in various traffic conditions. • HEVs can not save much fuel when the traffic is in free flow state. Despite extensive work on the fuel consumption of vehicles, the evaluation of hybrid electric vehicles (HEVs) requires a model that can predict fuel consumption second by second. In this paper, a fuel consumption model for hybrid vehicles is developed based on the operation of internal combustion engine (ICE), electric machine (EM) and battery. The fuel consumption model for engine is based on steady-state fuel mapping with a transient correction. EM and battery module (EBM) is introduced to estimate the fuel consumption for different modes of hybrid vehicle. The Argonne National Laboratory (ANL) measurements of different driving cycles including the steady-state cycle, the Urban Dynamometer Driving Schedule (UDDS) cycle, the Highway cycle and the United States 06 (US06) cycle are used to verified the model. With cellular automaton simulation, we study the fuel consumption in traffic flow and show that the fuel consumption can be significantly reduced by hybrid mode. The mean fuel saving is approximately 30% in the traffic flow on a horizontal road. The fuel consumption of traffic flow under uphill, downhill, tailwind and headwind conditions also showes considerable fuel saving. The model can be used to evaluate the fuel economy of HEVs, and provide eco-driving route-choice suggestions for drivers. [ABSTRACT FROM AUTHOR]

• Co-optimization of speed trajectory and power management is investigated. • Pontryagin's Minimum Principle is used to reduce the search space for control. • The performance of co- and sequential optimization is comprehensively analyzed. With the advent of vehicle connectivity via intelligent transportation systems, the simultaneous optimization of powertrain operation and vehicle dynamics has become an important research problem to improve energy efficiency. However, this problem is still very challenging due to computational loads in terms of time and memory usage, and hence very little effort has been spent to reveal the full potential of co-optimization. Studies in the literature on eco-driving or energy-efficient operation typically preferred using a sequential optimization or eliminating some constraints for fast computation. This paper extends the authors' previous work of determining energy-efficient speed profiles focused on vehicle dynamics by including additional state and control variables due to vehicle hybridization in application to a fuel-cell/battery electric vehicle. The results from Pontriagyn's Minimum Principle analysis reveal that the co-optimization can be formulated with one discrete variable describing vehicle operation and another continuous variable for power distribution to reduce computation in implementing Dynamic Programming. Then, the performance of co-optimization and sequential optimization for energy-efficient driving is comprehensively analyzed and compared in terms of energy consumption by each powertrain component, operating modes, and computational cost. The total energy savings by co-optimization range from 5.3% to 24.2% for aggressive driving on a hilly terrain. Meanwhile, on a relatively flat road, the benefit of co-optimization is less significant. Nonetheless, co-optimization achieved 0.5%–5.3% reduction in total energy consumption as compared to sequential optimization. [ABSTRACT FROM AUTHOR]